When Perplexity or ChatGPT answers with sources, a small set of pages wins the citations while everyone else gets paraphrased into anonymity. Watching citation patterns across engines in Sourceable, the winning pages are not the best-written — they are the most quotable. Citation is a formatting and evidence game, and it is learnable.
The uncomfortable prerequisite first: engines cite pages they can retrieve. If your best thinking is in a gated PDF or a JavaScript-only render, you have opted out before the game starts.
Content shapes that earn citations
Original numbers beat everything: benchmarks, survey results, "we analyzed N cases" — data that exists nowhere else forces attribution. After that: stated prices and ranges (engines love concrete answers to "how much" questions and few vendors publish them); genuine comparisons with criteria; definitions written to be lifted whole ("X is Y" in the first sentence); and dated, versioned claims — "in 2026, typical costs run..." — which signal freshness retrieval engines prefer.
“Engines do not cite the smartest page. They cite the page that committed to an answer and showed its evidence.”
Shapes that never get cited
Opinion essays without evidence — paraphrased, not quoted. "It depends" hedging — the model wants an answer, and cites whoever commits to one. Content that buries the claim in paragraph nine. And me-too explainers restating what fifty higher-authority pages already say: retrieval had better options before you published. If a page adds no fact the web did not already have, it will collect no citations.
The publishing habits that compound
Answer in the first 100 words, then earn the depth. One page per question — fragments dilute, consolidation wins. Refresh dated claims on a schedule, because engines demote stale numbers. Mark the page up with structured data so the machine parses what it is quoting. And publish on a cadence: citation share builds like compound interest, not like a campaign.
The pattern behind every tactic here: be the primary source for something. Our cost guides state real ranges for exactly this reason. The next piece covers the structured-data layer that makes all of it machine-legible.
Emperor Brains Engineering
Emperor Brains LLP
